School of Mathematical Sciences
Queen Mary, University of London
SUMMER 2001
STATISTICS SEMINAR: DESIGN OF EXPERIMENTS
All are welcome
The talks are held at 16.30, all in the Mathematics Seminar Room (103)
on Level 1 of the Mathematics Building, Queen Mary, University of London.
Tea and coffee are available in the Mathematics Common Room (102)
from 15.00.
The nearest underground station is Stepney Green.
Turn left at the exit and walk 400 yards.
________________________________________________________________________
DATE SPEAKER TITLE
------------------------------------------------------------------------
17 May 2001 Carola Deppe Crossover Designs for
De Montfort University Descriptive Analysis in Sensory
Leicester Evaluation
24 May 2001 Robert Mee Efficient Two-Level Designs for
University of Tennessee Estimating Main Effects
and Two-Factor Interactions
31 May 2001 Polly Martin Screening Large Numbers
University of Reading of Ingredients in the Development
of Agrochemical Formulations
7 June 2001 Martina Vandebroek The Design of Blocked and
Peter Goos Split-Plot Experiments
Katholieke Universiteit
Leuven
-----------------------------------------------------------------------
For more information ask:
Barbara Bogacka
School of Mathematical Sciences
Queen Mary, University of London
Mile End Road
London E1 4NS
Tel: 020 7882 5497
e-mail: [log in to unmask]
---------------------------------------------
The seminar information is kept on:
http://www.maths.qmw.ac.uk/~rab/seminars.html
______________________________________________________________________
A B S T R A C T S
______________________________________________________________________
Carola Deppe
"Crossover Designs for Descriptive Analysis in Sensory Evaluation"
One of the key aims of sensory analysis is to provide methods for the
objective assessment and comparison of product properties, based upon
responses derived from a panel of trained assessors. In the sensory
context, the choice of the statistical design for such studies takes on
additional significance in attempting to cope with the inherent
limitations of human response data. These limitations are especially
evident when large numbers of products need to be compared in a given
study, as is frequently the case, particularly within industry.
In this talk I will describe a three-step procedure for creating
crossover designs for multi-session experiments that have an additional
constraint on the number of products that can be prepared for a single
session.
----------------------------------------------------------------------
Robert Mee
"Efficient Two-Level Designs for Estimating Main Effects
and Two-Factor Interactions"
For seven or more factors, orthogonal resolution V designs require 2-4
times as many runs as there are main effects and two-factor interactions
to estimate.
By forfeiting orthogonality, it is possible to substantially reduce the
size of the design for estimating these linear and bilinear coefficients.
I will survey the literature on irregular resolution V designs and then
propose a class of designs that are fully efficient for main effects.
All bilinear coefficients can be estimated from these new designs,
though not with full precision.
------------------------------------------------------------------------
Polly Martin
"Screening Large Numbers of Ingredients in the Development
of Agrochemical Formulations"
The development of an agrochemical formulation generally involved
experimentation to optimise the efficacy and physical properties of the
formulation. In some cases it is desirable to test a large number of
potential ingredients in order to obtain the best subset for inclusion in
the final product.
When screening a large number of ingredients it can be useful to know,
not only the effect of each individual component on the measured response,
but also if an improved response can be obtained when two or more
components are combined.
Here a method is suggested for designing experiments to screen large
numbers of components and a practical example illustrates the use of such
designs.
-------------------------------------------------------------------------
Martina Vandebroek and Peter Goos
"The Design of Blocked and Split-Plot Experiments"
Although blocked and split-plot experiments are extremely popular in
practice, their optimal design has not received much attention in the
literature. In the presentation, we will describe the differences and the
similarities between blocked and split-plot experiments and propose a method
to compute the best possible tailor-made design in a particular experimental
situation. Examples will illustrate the computational results, the most
striking of which is that split-plot designs are sometimes more efficient
than completely randomized designs.
--------------------------------------------------------------------------
|